Following my talk this last January at Stanford Law School on the topic of AiCE, I was asked to expand my thoughts on the issue of AiCE rights. They were not the usual “what kind of rights should AI get” based questions, but rather much more basic, yet complex, multi-layered “why should we give these AI ‘things’ rights in the first place” types.

Thus, one of the central themes in the AiCE project addresses the issue of why such rights are necessary and also identifies which types (if any) are needed by AiCE for accomplishing its configured mission. The Uniform AiCE Transactions Act chapter, for example, introduces the concept of dynamic rights (DR). It prescribes that since an AiCE is multi-mission configurable, DR is a necessary and appropriate attribute that determines liability and scopes the available remedies when something goes wrong.

Complementary to DR, the AiCE project distinguishes between AiCE tasks that are inherently (and currently) the domain of natural persons and those that can be, from a normative perspective, comfortably performed by non-humans. The former is referred to as Human Centric Tasks (HCT), an example of which could be Professor Lawrence Solum’s “trustee” application and the latter is identified as Entity Neutral Tasks (ENT), such as CALO’s Siri. This differentiation disposes of the difficult reflexive “human-centric entity rights” analysis deployed whenever AI becomes the topic of conversation. This distinction is also in sync with the suprapsychological AI mode of analysis (identified by Professor Owen Flanagan), which broadens the “I” of AI to include the non-human variety.

In contrast to HCT, ENT configurations do not involve taking decisions that exert moral choice. These are, therefore, dispositive of non-malleable inquiries into liability, punishment, deterrence, and so forth. We can thus see that the HCT/ENT distinction serves to delineate as to the circumstances in which AiCE rights analysis is relevant. It also answers the question posed above (if only partially so) by stating that: (1) Not all AiCE/AI configurations will receive any rights; (2) only AiCE/AI that require rights for performance of their mission will receive them and (3) which rights AiCE/AI receive will depend on the type of task they are configured to carry out.